fastai makes deep learning with PyTorch faster, more accurate, and easier
The fastai deep learning library. See the fastai website to get started.
Note for course.fast.ai students
If you are using fastai for any course.fast.ai course, please do NOT install fastai from pip or conda using the instructions below; the instructions below are for fastai v1, but the courses use fastai 0.7. For the courses, you should simply follow the instructions in the course (i.e. clone this repo, cd to it, and
conda env update), and the notebooks will work (there is a symlink to old/fastai/, which is fastai 0.7, in each course notebook directory).
Is My System Supported?
Python: You need to have python 3.6 or higher
Since fastai-1.0 relies on pytorch-1.0, you need to be able to install pytorch-1.0 first.
As of this moment pytorch.org's pre-1.0.0 version (
- linux: fully
- mac: CPU-only
- windows: not supported
This will change once
pytorch1.0.0 is released and installable packages made available for your system, which could take some time after the official release is made. Please watch for updates here.
If your system is currently not supported, please consider installing and using the very solid "v0" version of
fastai. Please see the instructions.
To install fastai with pytorch-nightly + CUDA 9.2 simply run:
conda install -c pytorch -c fastai fastai pytorch-nightly cuda92
If your setup doesn't have CUDA support remove the
cuda92 above (in which case you'll only be able to train on CPU, not GPU, which will be much slower). For different versions of the CUDA toolkit, you'll need to install the appropriate CUDA conda package based on what you've got installed on your system (i.e. instead of
cuda92 in the above, pick the appropriate option for whichever toolkit version you have installed; to see a list of options type:
conda search "cuda*" -c pytorch).
NB: We are currently using a re-packaged torchvision in order to support pytorch-nightly, which is required for using fastai.
If your system doesn't have CUDA, you can install the CPU-only torch build:
conda install -c pytorch -c fastai fastai pytorch-nightly==1.0.0.dev20180928=py3.6_cpu_0
First install the nightly
pytorch build, e.g. for CUDA 9.2:
pip install torch_nightly -f https://download.pytorch.org/whl/nightly/cu92/torch_nightly.html
If you have a different CUDA version find the right build here. Choose Preview/Linux/Pip/python3.6|python3.7 and your CUDA version and it will give you the correct install instruction.
Next, install a custom
torchvision build, that is built against
pip install --index-url https://test.pypi.org/simple/ --extra-index-url https://pypi.org/simple/ torchvision==0.2.1.post1
Now you can install
fastai. Note, that this is a beta test version at the moment, please report any issues:
pip install fastai
Sometimes, the last
pip command still tries to get
torch-0.4.1. If that happens to you, do:
pip uninstall torchvision fastai pip install --no-deps torchvision pip install fastai
First, follow the instructions above for either
Conda. Then remove the fastai package (
pip uninstall fastai or
conda uninstall fastai) and replace it with a pip editable install:
git clone https://github.com/fastai/fastai cd fastai tools/run-after-git-clone pip install -e . pip install jupyter_contrib_nbextensions ipywidgets
You can test that the build works:
jupyter nbconvert --execute --ExecutePreprocessor.timeout=600 --to notebook examples/tabular.ipynb
Copyright 2017 onwards, fast.ai, Inc. Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. A copy of the License is provided in the LICENSE file in this repository.
- Released on Conda and Pypi
- First release on PyPI.
Release history Release notifications | RSS feed
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
|Filename, size||File type||Python version||Upload date||Hashes|
|Filename, size fastai-1.0.3-py3-none-any.whl (87.6 kB)||File type Wheel||Python version py3||Upload date||Hashes View|
|Filename, size fastai-1.0.3.tar.gz (78.8 kB)||File type Source||Python version None||Upload date||Hashes View|